State of health monitoring in machining: Extended Kalman filter for tool wear assessment in turning of IN718 hard-to-machine alloy

An extended Kalman filter (EKF) is employed in this work for tracking tool flank wear area in wet-turning of Inconel 718 (IN718) nickel-based alloy in variable feed condition. The tool wear area evolution is modeled with a 3rd order polynomial empirical function and an analytical solution for discrete state space system is derived. The state uncertainty was found to decrease up to a critical range of 200-250 μm of average flank wear length, and then increase abruptly with an increase in tool wear. Therefore, the tool wear uncertainty was modeled with failure probability density, i.e. the bathtub function, while a constant uncertainty was considered for the measurement signal (spindle power). To demonstrate the significance of using this method, the root mean square error (RMSE) and the mean absolute error (MAE) were calculated and compared with deterministic method in estimation of the tool wear area. It was shown that the proposed estimation based on stochastic filter EKF increased the accuracy of estimation by maximum of 60%. Results for estimation of the rate of tool wear area indicate additional possible effects or transitions of effects at higher wear conditions.